reproduce the LLaVA-NeXT-Video 7B scores

[junho-jo]

reference: https://lmms-lab.github.io/posts/lmms-eval-0.2/,

In the section of "Alignment Check for Video Datasets", there are LLaVA-Next-Video-7B scores in the benchmark table.
Could you share the corresponding model checkpoint link? I'm very confused with the naming of version..

1. https://huggingface.co/lmms-lab/LLaVA-NeXT-Video-7B
2. https://huggingface.co/lmms-lab/LLaVA-Video-7B-Qwen2

And also the scripts for evaluation?

[giobin]

i have the same issue. Difficult to understand exactly which checkpoint and model_name to use to reproduce those results